B15-08 Patient Stratification Using Serum Cytokines in Acute Exacerbations of Chronic Obstructive Pulmonary Disease: Relationship to Clinical Manifestations and Outcomes

Introduction Heterogeneity of the host response within acute exacerbations of chronic obstructive pulmonary disease (AECOPD) limits discovery and targeting of immunomodulatory therapies. Clustering approaches using sputum biomarkers have defined distinct inflammatory subendotypes in AECOPD. It is unknown if similar subendotypes could be seen in peripheral blood and how these subendotypes related with disease severity and clinical outcomes. We investigated whether inflammatory clusters based on circulating cytokine protein were seen in AECOPD, and the relationships with clinical characteristics and outcomes. Methods We analyzed data from the AECOPD inpatient registry study (ACURE), a nationwide, multicenter cohort of prospectively enrolled patients hospitalized with AECOPD in China. We randomly selected 512 patients from the original cohort, measured 22 serum cytokines and related proteins from blood samples collected at admission, and used unsupervised hierarchical clustering to identify subphenotypes. Patient clinical characteristics and outcomes during hospitalization were compared after subendotype assignment. A parsimonious model using multinomial LASSO and logistic regression was developed to classify patient subendotypes. Results Overall, hierarchical cluster analysis best described the population by three clusters based on inspection of the dendrograms, also favored by k-means and consensus clustering. Cluster 1 (21.7%) comprised patients characterized by significantly higher concentrations of type 2 (T2) cytokines [e.g., interleukin-4 (IL-4), IL-5, and Eotaxin-1], termed as a T2 inflammation-predominant subphenotype. Cluster 2 (24.4%) exhibited significantly elevated levels of T1 and T3 cytokines [e.g., IL-8, tumor necrosis factor-1 (TNFR-1), TNFR-2 and IL-2], termed as a T1/T3 subendotype. Cluster 3 (53.9%) exhibited levels of all cytokines significantly below the overall mean, termed as a pauci-inflammatory subphenotype (Figure 1). The inflammatory endotype is associated with distinct clinical characteristics and outcomes, with T2 showing the highest median eosinophil counts (0.14*10^9/L, p = 0.016) and T1/T3 exhibiting the highest median neutrophil counts (4.81*10^9/L, p = 0.037). T2 also showed features of less severe disease and better prognosis during admission than the other clusters, such as the lowest mMRC score (2, median, p = 0.016), the shortest hospital stays (p = 0.021), and reduced duration of antibiotic and corticosteroid use (p = 0.030; p = 0.024). A parsimonious three-biomarker model (IL2, IL-4, IL-5) identified subendotypes in an internal validation cohort (T2: AUC 0.99; T1/T3: AUC 0.91; Pauci: AUC 0.96). Conclusion These findings demonstrate that differences in response at the level of serum cytokine biology show clustering in AECOPD patients, being related to disease severity and clinical outcome. Patient subendotype assignment which was characterized by distinct biological profiles could inform contextualized future study design. This abstract is funded by: None